부산대학교 도서관

A Fast and Lossless Image Compression (FLIC) algorithm based on the median edge predictor and Golomb coder of JPEG-LS is presented. FLIC eliminates the gradient-based context model from the JPEG-LS standard, the most expensive parts with respect to computational complexity and memory space requirements. To avoid a large context memory, Golomb parameter is selected based on the coding states and the prediction residuals of up to two immediate neighbors, one in each dimension. The FLIC algorithm has low memory footprint and dissolves the data dependencies in JPEG-LS to facilitate parallelization. Experimental results show that the FLIC algorithm achieves a throughput speedup factor of 3.7 over JPEG-LS with less than 4% compression performance penalty. Lossless compression performance results further show that FLIC outperforms other state-of-the-art standards including JPEG 2000 and JPEG XR.